Is the order in which a UNION ALL query returns rows guaranteed?

Question

Guys
I'm doing a kind of query where I want to return matching rows in a priority order and take the first one. THink of it as getting the most specific error message for a particular context.
SELECT * FROM (
SELECT msg FROM errormessages WHERE device = 'x'
UNION ALL
SELECT msg FROM errormessages WHERE client = 'y'
UNION ALL
SELECT msg FROM (SELECT msg from errormessages WHERE class LIKE 'generic_' ORDER BY class)
UNION ALL
SELECT 'missing error message' FROM dual
WHERE ROWNUM =1
If UNION ALL is guaranteed to return rows in this order, great. If not, would it be better to have:
SELECT * FROM ( SELECT * FROM(
SELECT 1 as ord, msg FROM errormessages WHERE device = 'x'
UNION ALL
SELECT 2, msg FROM errormessages WHERE client = 'y'
UNION ALL
SELECT 4+SUBSTR(class, 7, 1), msg FROM errormessages WHERE class LIKE 'generic_'
UNION ALL
SELECT 9999, 'missing error message' FROM dual
) ORDER BY ord )
WHERE ROWNUM =1Or, another option:
SELECT msg FROM(
SELECT
  CASE
    WHEN device = 'x' THEN 1
    WHEN client = 'y' THEN 2
    WHEN class LIKE 'generic_' THEN 4+SUBSTR(class, 7, 1)
  END as ord, msg
FROM (
  SELECT * FROM errormessages WHERE device = 'x' OR  client = 'y' OR class LIKE 'generic_'
  UNION ALL
  SELECT 999, 'missing error' FROM dual
) ORDER BY ord
)WHERE ROWNUM =1As you can see there is a complication in the generic; devices may advertise that they accept 1,2,3 or 4 lines of info, so I want to use the advertised capabiltiy to pick the most relevant message from the generic set if no specifics exist for the device or the client
Which was would you go?
Cheers
Edited by: charred on Feb 11, 2009 4:56 AM - code tags

Answer 1

charred wrote:
So tell me guys, do I go for:
4 queries unioned together
or
1 query with 4 ORs in the where clause and a case when to determine priority?It will depend on your conditions, but if you look at your explain plans you should get an idea for which is the better one; typically I would expect this to be the OR method...
SQL> select * from emp where deptno = 20
  2  union all
  3  select * from emp where job = 'CLERK'
  4  union all
  5  select * from emp where sal > 2500;
Execution Plan
Plan hash value: 3153085224
| Id  | Operation                    | Name     | Rows  | Bytes | Cost (%CPU)| Time     |
|   0 | SELECT STATEMENT             |          |    14 |   546 |     8  (75)| 00:00:01 |
|   1 |  UNION-ALL                   |          |       |       |            |          |
|   2 |   TABLE ACCESS BY INDEX ROWID| EMP      |     5 |   195 |     2   (0)| 00:00:01 |
|*  3 |    INDEX RANGE SCAN          | DEPT_IDX |     5 |       |     1   (0)| 00:00:01 |
|*  4 |   TABLE ACCESS FULL          | EMP      |     4 |   156 |     3   (0)| 00:00:01 |
|*  5 |   TABLE ACCESS FULL          | EMP      |     5 |   195 |     3   (0)| 00:00:01 |
Predicate Information (identified by operation id):
   3 - access("DEPTNO"=20)
   4 - filter("JOB"='CLERK')
   5 - filter("SAL">2500)
Statistics
          1  recursive calls
          0  db block gets
         18  consistent gets
          1  physical reads
          0  redo size
       1328  bytes sent via SQL*Net to client
        396  bytes received via SQL*Net from client
          2  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
         14  rows processed
SQL> select * from emp
  2  where deptno = 20 or job = 'CLERK' or sal > 2500;
Execution Plan
Plan hash value: 3956160932
| Id  | Operation         | Name | Rows  | Bytes | Cost (%CPU)| Time     |
|   0 | SELECT STATEMENT  |      |    10 |   390 |     3   (0)| 00:00:01 |
|*  1 |  TABLE ACCESS FULL| EMP  |    10 |   390 |     3   (0)| 00:00:01 |
Predicate Information (identified by operation id):
   1 - filter("DEPTNO"=20 OR "SAL">2500 OR "JOB"='CLERK')
Statistics
          1  recursive calls
          0  db block gets
          8  consistent gets
          0  physical reads
          0  redo size
       1146  bytes sent via SQL*Net to client
        396  bytes received via SQL*Net from client
          2  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          9  rows processed
SQL>